Implicit Geolocation of Social Networking Users

1. A method comprising, by one or more computing systems: receiving, from a requesting service, a request for a location prediction for a user; generating a current location prediction from a plurality of previous location signals, wherein each previous location signal comprises an indication of the user's location at a respective time prior to a predetermined time span, and wherein each previous location signal is weighted based on (1) a type of the previous location signal and (2) a difference between a current time and the respective time for the previous location signal; and sending, in response to the request, the current location prediction for the user to the requesting service.

2. The method of claim 1 , wherein the plurality of previous location signals comprise the user's IP address.

3. The method of claim 1 , wherein the plurality of previous location signals comprise a self-declared user location.

4. The method of claim 1 , wherein the plurality of previous location signals comprise real-time location signals received before the predetermined time span.

5. The method of claim 1 , wherein generating the current location prediction further comprises weighting the plurality of previous location signals based on a relative reliability of each previous location signal.

6. The method of claim 1 , further comprising storing, when each previous location signal is received, each previous location signal in a histogram, wherein each previous location signal is weighted based on predetermined time decay for the histogram.

7. The method of claim 1 , further comprising storing, when each previous location signal is received, each previous location signal in one of a fast decay histogram or a slow decay histogram, wherein the fast-decay histogram comprises previous location signals of a first type, and the slow-decay histogram comprises previous location signals of a second type.

8. The method of claim 1 , further comprising: generating a plurality of location predictions; and filtering the plurality of location predictions based on a set of rules to generate the current location prediction, wherein the set of rules comprises filtering out location predictions that exceed a predetermined geographic radius from a location associated with the user's IP address.

9. The method of claim 1 , wherein generating the current location prediction comprises calculating a weighted average for a plurality of location predictions, wherein a weight of each prediction is based at least in part on a source of signals from which the prediction was generated.

10. The method of claim 1 , wherein generating the current location prediction comprises weighting each previous location signal, and wherein the weight applied to each previous location signals is determined via a machine learning algorithm.

11. One or more computer-readable non-transitory storage media embodying software that is operable when executed to: receive, from a requesting service, a request for a location prediction for a user; generate a current location prediction from a plurality of previous location signals, wherein each previous location signal comprises an indication of the user's location at a respective time prior to a predetermined time span, and wherein each previous location signal is weighted based on (1) a type of the previous location signal, and (2) a difference between a current time and the respective time for the previous location signal; and send, in response to the request, the current location prediction for the user to the requesting service.

12. The media of claim 11 , wherein the plurality of previous location signals comprise the user's IP address.

13. The media of claim 11 , wherein the plurality of previous location signals comprise a self-declared user location.

14. The media of claim 11 , wherein the plurality of previous location signals comprise real-time location signals received before the predetermined time span.

15. The media of claim 11 , wherein generating the current location prediction further comprises weighting the plurality of previous location signals based on a relative reliability of each previous location signal.

16. The media of claim 11 , wherein the software is further operable to store, when each previous location signal is received, each previous location signal in a histogram, wherein each previous location signal is weighted based on predetermined time decay for the histogram.

17. The media of claim 11 , wherein the software is further operable to store, when each previous location signal is received, each previous location signal in one of a fast decay histogram or a slow decay histogram, wherein the fast-decay histogram comprises previous location signals of a first type, and the slow-decay histogram comprises previous location signals of a second type.

18. The media of claim 11 , wherein the software is further operable to: generate a plurality of location predictions; and filter the plurality of location predictions based on a set of rules to generate the current location prediction, wherein the set of rules comprises filtering out location predictions that exceed a predetermined geographic radius from a location associated with the user's IP address.

19. The media of claim 11 , wherein generating the current location prediction comprises calculating a weighted average for a plurality of location predictions, wherein a weight of each prediction is based at least in part on a source of signals from which the prediction was generated.

20. A system comprising: one or more processors; and a memory coupled to the processors comprising instructions executable by the processors, the processors being operable when executing the instructions to: receive, from a requesting service, a request for a location prediction for a user; generate a current location prediction from a plurality of previous location signals, wherein each previous location signal comprises an indication of the user's location at a respective time prior to a predetermined time span, and wherein each previous location signal is weighted based on (1) a type of the previous location signal, and (2) a difference between a current time and the respective time for the previous location signal; and send, in response to the request, the current location prediction for the user to the requesting service.